Flavours and Fragrances

13.2 Terpenes as Renewable Resources for Terpene Flavour Molecules 293 13.2.5.2 Hotrienol from Linalool Hotrienol was found for the first time in Ho leaf oil as the S enantiomer [7], but has been found since then in many natural sources; for instance, the R enantio- mer was found in black tea and in green tea. The product can be used in many flavours, such as elderflower, grape, berry and honey flavours. It can be prepared from linalool obtained from citrus oils or Chinese Ho oils, but most linalool is obtained by synthesis from isoprene from petrochemical sources. Recently a practical and convenient synthesis was described starting from linalool via linalyl acetate [8]. It involves the ene-type chlorination of linalyl ac- etate prepared from linalool which results in the formation of γ-chloro-α-linalyl acetate (Scheme 13.8). Dehydrochloronation with lithium bromide and lithium carbonate in dimethylformadide followed by hydrolysis of dehydro-α-linalyl ac- ylate results in hotrienol. Scheme 13.8 Chemical synthesis of hotrienol from linalool 13.2.5.3 Nootkatone from Valencene Nootkatone is an important constituent from grapefruit flavour. It is synthesised by oxidation of valencene, which is obtained and isolated from orange peel oil where it occurs at a maximum level of 0.4% [9]. For the oxidation of valencene to nootkatone, strong oxidation agents like oxygen in the presence of metal salts, peroxides or chromate compounds are required. There are also several patents on the bioconversion of valencene to nootkatone which results in natural nootkatone (Scheme 13.9) [10, 11]. Scheme 13.9 Conversion of valencene into nootkatone

294 13 Chemical Conversions of Natural Precursors 13.2.5.4 Terpene Esters The most important and frequently used terpene esters in flavours are the ac- etates of nerol, geraniol, citronellol, linalool and isoborneol [12]. As discussed before, all these terpene alcohols are available both from renewable resources and from petrochemical origin. Acetic acid can be obtained from renewable re- sources by pyrolysis of wood as wood vinegar, and also by synthesis from pet- rochemical origin. 13.3 Vanillin 13.3.1 Vanillin Synthesis Eugenol obtained from clove oil is an important precursor for the preparation of vanillin (Scheme 13.10). The reaction consists of two steps. First, eugenol needs to be converted into isoeugenol, which requires alkaline treatment or ruthe- nium or rhodium catalysis. Second, the isoeugenol is oxidised to vanillin using, for instance, chromic acid. This method results in nature-identical vanillin. Lignin waste from the wood pulp industry also serves as a renewable source of precursors for the preparation of vanillin. Lignin can be degraded by treat- ment with alkali and oxidising agents to coniferyl alcohol like structures which can be oxidised to vanillin. Curcumin, a yellow colouring material from the roots of Curcuma longa, can also serve as a precursor for vanillin; from each molecule of curcumin two molecules of vanillin can be derived. Also a lot of synthetic vanillin is prepared from petrochemical sources such as phenol, cat- echol and guaiacol. 13.3.2 Vanillin Derivatives Vanilla flavour is not only determined and characterised by the vanillin mol- ecule, but also by many more phenolic compounds and vanillin derivatives. Two examples of molecules that recently obtained FEMA-GRAS status are vanillyl ethyl ether and vanillin 2,3-butanediol acetal (Scheme 13.11). Vanillin can be hydrogenated to form vanillyl alcohol, which is also used in vanilla flavours. Vanillyl alcohol can be reacted with ethanol to form vanillyl ethyl ether. Vanillin can also form an acetal with 2,3-butanediol (obtained by fermentation of sug- ars) catalysed by p-toluene sulfonic acid in toluene.

13.3 Vanillin 295 Scheme 13.10 Chemical synthesis of vanillin from various sources 13.3.3 Heliotropine from Safrole Safrole is used for the preparation of heliotropine, which is mainly used in va- nilla flavours. Safrole has to be converted into isosafrole by alkali treatment or ruthenium or rhodium catalysis analogous to the eugenol to isoeugenol conver- sion before it can be oxidised to heliotropine using chromic acid as a catalyst [13] (Scheme 13.12). Safrole, which is both toxic and carcinogenic, occurs in sassafras oil up to 90%. There are two commercially important sassafras oils: the Brazilian oil is obtained from the trunk wood of Ocotea pretiosa and the Chinese oil is obtained from Cinnamomum camphora by steam distillation of wood chips. Safrole from

296 13 Chemical Conversions of Natural Precursors Scheme 13.11 Formation of vanillin ethyl ether and vanillin 2,3-butanediol acetal from vanillin sassafras oil is not only used to prepare heliotropine, but also for the synthesis of piperonal butoxide, which is a vital ingredient for pyrethroid insecticides. The demand for sassafras oil is 2,000 t. Brazil is confronted with a depletion of its natural resource for sassafras oil, because more trees are felled than are replanted and mature [14]. This has resulted in measures involving restrictions on felling trees, resulting in a production decline of sassafras oil. This illustrates the fact that also renewable resources can be depleted and result in a shortage, which can be a long-term problem, especially when the wood from full-grown trees is required. Currently Piper species, which are weeds or shrubs indigenous in Central and South America, are being investigated as an alternative source of safrole. Propagation studies and growing trials have shown that these plants may act as renewable resources when cultivated on plantations. Biomass yields and oil productivity per hectare per year together with projected oil prices will determine whether economic returns to the farmer can be guaranteed. Scheme 13.12 Conversion of safrole into heliotropine

13.4 Sugars as Precursors 297 13.4 Sugars as Precursors 13.4.1 Sources of Xylose and Rhamnose Sugars are important as renewable resources for the preparation of process fla- vours. On the other hand, some rare sugars like xylose and rhamnose are also important for the preparation of specific flavour chemicals. The C5 sugar xylose can be obtained from plant materials like wood, straw and hulls; acid hydrolysis of its precursor xylan, which is a polysaccharide built up from D-xylose, results in xylose. The C6 deoxy sugar rhamnose also occurs in many plants. Commer- cially available rhamnose is obtained by chemical hydrolysis of arabic and ka- raya gums, or from rutin (the rhamnoglucoside of quercetin), which is present in many plants. Also naringin, which is the main bitter compound in grapefruit, occurs in the rind of citrus fruits and is a source of rhamnose. 13.4.2 Examples of Flavour Chemicals Derived from Sugars 13.4.2.1 2-Methylfuran-3-thiol from Xylose and Hydrogen Sulfide 2-Methylfuran-3-thiol is responsible for a boiled note in milk, chicken and beef flavours and this chemical as well as its disulfide are widely used in flavours. Both molecules can be prepared by heating xylose and hydrogen sulfide in an autoclave (Scheme 13.13). Alternatively, these sulfur compounds can also be synthesised from 2-methylfuran made from furfural derived from oat hulls. Scheme 13.13 Formation of 2-methylfuran-3-thiol and its disulfide from xylose 13.4.2.2 4-Hydroxy-2,5-dimethyl-3(2H)-furanone from Rhamnose 4-Hydroxy-2,5-dimethyl-3(2H)-furanone is a widely used flavour molecule for fruit and brown flavours which is prepared by heating of rhamnose. A nitrogen-

298 13 Chemical Conversions of Natural Precursors containing base like proline or piperidine serves as a catalyst for its formation (Scheme 13.14). Scheme 13.14 Formation of 4-hydroxy-2,5-dimethyl-(2H)-furanone from rhamnose 13.4.2.3 Pyrazines from Rhamnose and Ammonia Pyrazines are mainly used in roasted, peanut and chocolate flavours. Alkyl pyr- azines can be obtained by the reaction in an autoclave at high temperature (120– 160 °C) of reducing sugars like rhamnose and ammonia. 13.4.2.4 Furfural As a Precursor for Furfuryl Mercaptan Furfural comes from pentose sugars in cereal straws and brans. Furfural is the precursor of furfuryl mercaptan and its disulfide, difurfuryl disulfide, which are both important chemicals for coffee, meat and roasted flavours. They are pre- pared by the reaction of furfural and hydrogen sulfide (Scheme 13.15). Scheme 13.15 Formation of furfurylthiol and its disulfide from furfural

13.5 L-Cysteine and L-Methionine as Sources of Hydrogen Sulfide and Methanethiol 299 13.5 L-Cysteine and L-Methionine as Sources of Hydrogen Sulfide and Methanethiol 13.5.1 Cysteine Cysteine can be obtained by hydrolysis from cysteine-rich proteins in hair or feathers or from petrochemical sources. Cysteine is an important raw material in Maillard reactions for the preparation of process flavours, but it can also serve as a source of ammonia and hydrogen sulfide for the preparation of flavour chemicals, such as the terpene sulfur compounds mentioned in Sect. 13.2.4 and furfuryl mercaptan mentioned in Sect. 13.4.2.4. 13.5.2 Methionine Methionine can be obtained from enzymatic protein hydrolysates or from pet- rochemical sources. To a lesser extent than cysteine, it is a raw material in Mail- lard reactions for the preparation of process flavours and it can also be utilised as a precursor for the chemical preparation of the sulfide methional, which is an important flavour constituent for potato, malt, seafood and many other flavours. Methional can be reduced to methionol, which can be esterified with organic acids to, for instance, methionyl acetate and methionyl butyrate, which are use- ful compounds for pineapple and other fruit flavours (Scheme 13.16). Scheme 13.16 Methionine as a source of flavour chemicals 13.6 Chemical Conversions of Natural Precursors Obtained by Fermentation or from Residual Streams 13.6.1 Aliphatic and Aromatic Esters A very important group of flavour molecules is the esters. Many flavour esters can be prepared from organic acids and alcohols from renewable resources. Im-

300 13 Chemical Conversions of Natural Precursors portant alcohols readily available from renewable resources are methanol from wood pyrolysis, ethanol and butanol by fermentation of sugars, propanol and (iso)amyl alcohol from fusel oil, and also benzyl and phenyl ethyl alcohol from essential oils. Many of these alcohols are also prepared in large quantities from petrochemical sources. The same applies to organic acids like acetic acid from wood pyrolysis, lactic and butyric acids from fermentation of sugars, hexanoic, octanoic, decanoic, myristic and oleic acids from vegetable oils and levulinic acid from cellulose; most of these acids also have economical petrochemical sources. 13.6.2 Heterocyclic Flavour Molecules Also heterocyclic flavour molecules can be formed from renewable resources. 3,5-Diethyl-1,2,4-trithiolane is an important molecule for onion flavours and can easily be prepared from propanal obtained by biotransformation and hy- drogen sulfide (Scheme 13.17). A meat flavour molecule like thialdine [dihydro- 2,4,6-trimethyl-1,3,5(4H)-dithiazine] can be prepared from acetaldehyde iso- lated from molasses and ammonium sulfide (Scheme 13.18). The bacon flavour substance 2,4,6-triisobutyl-5,6-dihydro-4H-1,3,5-dithiazine can be prepared from isovaleraldehyde prepared from essential oils and ammonium sulfide (Scheme 13.19). Scheme 13.17 Formation of 3,5-diethyl-1,2,4-trithiolane from acetaldehyde Scheme 13.18 Formation of thialdine from acetaldehyde Scheme 13.19 Formation of 2,4,6-triisobutyl-5,6-dihydro-4H-1,3,5-dithiazine from isovaleralde- hyde

13.7 Conclusions 301 13.7 Conclusions Many products from the flavour industry are primary products from renewable resources or secondary products obtained by chemical conversions of the pri- mary products. In general these secondary products are key flavour chemicals with a high added value. The cost difference between a precursor, the primary product and the flavour chemical can easily amount to a factor 20–1,000, es- pecially when it concerns a natural flavour chemical. A large part of this cost reflects, of course, the efficiency of the reaction, the labour involved and the cost of the other reagents. Although quite often these flavour chemicals can be prepared from petro- chemical sources, renewable resources are preferred by the flavour industry, be- cause access to these renewable resources is very good and already existed when these companies were started. In addition, chemicals from renewable resources are natural, so they can be used in natural flavours and offer the possibility to be used for the production of natural secondary products. References 1. Lawrence BM (1993) In: Janick J, Simon JE (eds) New crops. Wiley, New York, p 620 2. Sanganeria S (2005) Perfum Flavor 30(7):24 3. Zaobang S (1995) CIFOR occasional paper 6. Center for International Forestry Research, Bogor 4. Sell CS (2003) A fragrant introduction to terpenoid chemistry. Royal Society of Chemistry, Cambridge 5. Demole E, Enggist P, Ohloff G (1982) Helv Chim Acta 65:1785 6. Lamparsky D, Schudel P (1971) Tetrahedron Lett 36:3323 7. Yoshida T, Muraki S, Kawamura H, Komatsu A (1969) Agric Biol Chem 33:343 8. Yuasa Y, Kato Y (2003) J Agric Food Chem 51:4036 9. MacLeod AJ, MacLeod G, Subramanian G (1988) Phytochemistry 27:2185 10. Muller B, Dean C, Schmidt C, Kuhn JC (1997) WO 9722575 11. Huang R, Christensen PA, Labuda IM (2001) US Patent 6,200,786 12. Wright J (2004) Flavor creation. Allured, Carol Stream 13. Dorsky J (1991) In: Mueller PM, Lamparsky D (eds) Perfumes: art, science and technology. Elsevier, Amsterdam, chap 14 14. Coppen J (1995) Flavors and fragrances of plant origin. FAO, Rome

14 Industrial Quality Control Herbert J. Buckenhueskes LWB—Lebensmittelwissenschaftliche Beratung, Hirschstrasse 25, 71282 Hemmingen, Germany 14.1 Introduction During the last two decades the term “quality” has become one of the most stressed words in the field of food and food production. The facts behind this are, on the one hand, the traditionally different meanings of the word “qual- ity” and, on the other hand, the advanced importance of quality and quality management systems as tools for an economical and safe production of food. “Quality” originates from the Latin language meaning as much as “property” or “characteristic”. In relation to food it originally was used as a synonym for “freshness” and “unspoilt”. From antiquity up to now, many philosophers, scientists and economists have tackled with the sense and the meaning of the term “quality”, leading to numer- ous quality models, for example metaphysical, product management, economi- cal, ecological and cognitive approaches. To get an overview of the major quality models, see [1]. In the book Flavor Science—Sensible Principles and Techniques, Acree and Teranishi [2] distinguished two different meanings of the word “quality”. One meaning is that of an attribute, for example sweet, bitter or floral. The second meaning depends on whether someone likes these attributes, by which quality relates to acceptance and the question how people interact with it. 14.2 Quality and Quality Management Systems According to the International Organisation for Standardisation (ISO) “quality” is the entirety of attributes and characteristics of a product or a service which are necessary to fulfil its defined or assumed requirements. As already indicated, quality can be seen to be more or less comprehensive, for which reason many organisations define or describe their commitment to quality in a so-called quality policy statement. A quality policy typically is based on three fundamen- tal principles:

304 14 Industrial Quality Control 1. Ensuring that the customer’s needs are identified and that these are con- formed. 2. Examination of all production and service processes in order to identify the potential for errors and to take necessary actions to eliminate them. 3. Ensuring that each employee understands how to do his/her job and is doing it right. In order to implement the quality policy in the daily work, quality manage- ment systems are installed, covering quality planning, quality control, quality as- surance and quality improvement. To ensure that the quality assurance system is in place and effective, external standards are used, for example the DIN EN ISO 9000 ff. standard system, commonly shortened to ISO 9000 (DIN is an acronym for Deutsches Institut für Normung, meaning “German Industry Standard”). The best known international quality management standard seems to be the so-called DIN EN ISO 9001 ff. standard. ISO standards are agreements devel- oped by technical committees. Since the members of these committees come from many countries, ISO standards tend to have very broad support. Confor- mance to that is said to guarantee that a company provides quality services and products. This standard was amended in 2000, so the current standard is called DIN EN ISO 9000:2000. The most important steps to follow the ISO 9000 standard are: • Deciding quality assurance policies and objectives • Formally writing down the company’s policies and requirements and how the staff can implement the quality assurance system • Implementation of the quality assurance system • Examination of the quality assurance system by an outside assessor to see whether it complies with the ISO standard • Describing the parts of the standard the company is missing and correction of any problem • Certification that the company is in conformance with the standard Beside the ISO standards there are some other standards which are set up by different organisations and sometimes it is really a problem to fulfil the require- ments of the different partners in the food market. Not least because of this situation it is not possible here to go in more details. For further information on quality systems and quality management systems see the specific standards as well as the respective literature. Very special quality demands and quality systems are those for the produc- tion and certification of kosher and halal products, which at the time are gaining in importance all over the world. Kashrut is the body of Jewish law dealing with what foods Jews can and cannot eat and how those foods must be prepared. “Kashrut” originates from the Hebrew and means “fit”, “proper” or “correct”. The more commonly known word “kosher” comes from the same roots and refers to foodstuffs that meet these dietary requirements of Jewish law. “Halal” is an

14.3 Quality Control 305 Arabic word meaning “lawful” or “permitted”, and eating halal is obligatory for every Muslim. The opposite of “halal” is haram, which means “prohibited”. Whether a company fulfils the requirements for a kosher or halal production or not can be examined and certificated by specially qualified people or organi- sations. For further information, see the recently published books concerning kosher [3] and halal [4] production. 14.3 Quality Control In the frame of a quality management system, quality control is defined as a set of activities or techniques whose purpose is to ensure that all quality require- ments are being met. Every raw material used, all intermediate products as well as all flavours and flavouring products which are delivered to the customers have to be controlled by appropriate physicochemical, biotechnological (e.g. enzymic or immunologic procedures), sensory or if necessary microbiological methods. The quality control of flavourings as well as their raw materials is a highly com- plex field and quality control laboratories in the flavour industry may have more than 500 defined analytical procedures [5]. A prerequisite for any quality control is the definition of how the character- istics of a specific raw material, an intermediate product or a final product of a manufacturing process should be described. This means that all characteristics for every single product have to be defined in adequate standards and specifi- cations so that the results obtained can be compared with these data. Numer- ous standards and specifications have been established in more or less official specification collections, for example pharmacopoeias, the aforementioned ISO or DIN standards, standards of the Essential Oil Association or the American Spice Trade Organization (ASTA). According to [6], the main objectives of the quality control in the flavour industry concern the following items: • Identity: Does the raw material delivered or the manufactured product cor- respond with the order? • Purity: Are the raw materials or the manufactured product free from unac- ceptable impurities, e.g. filth? • Contamination: Is there contamination, e.g. heavy metals, pesticides, aflatox- ins, microorganisms? • Adulterations: Are the raw materials free from adulterations? • Quantitatively limited substances: Are legal regulations concerning limited amounts of specific substances observed? • Spoilage: Ageing and unsuitable storage conditions can lead to quality changes of raw materials or products up to the complete spoilage of the product. • Authenticity: Conformity of the declared and real origin of a raw material. Are materials which are declared as “natural” really natural and not synthetic?

306 14 Industrial Quality Control The extensive quality control tests of raw materials, intermediate and final products represent a flood of data which have to be evaluated and documented according to the different aims of the quality control system. Considering the fact, that quality control often has to work under deadline pressure this work can only be done by using powerful electronic labour information and manage- ment systems (LIMS). 14.4 Physicochemical Methods Supported by the overall development in all fields of analysis during the past few decades, a precise analytical methodology has been developed for the dif- ferent aspects of quality control, comprising physicochemical, biotechnological, sensory and microbiological methods. In order to meet the sense of the qual- ity control system and by that the customers requirements, all methods applied have to be validated by adequate quality assurance tools. In the frame of this short review it is not even possible to discuss only the ma- jor methods and techniques used in industrial quality control in detail, so they will only be summarised here [5, 7]. For sample preparation, isolation and separation traditional methods like distillation (e.g. essential oil content of raw materials) or Soxhlet extraction are still in use. Beyond that, more recent methods are employed, for example super- critical fluid extraction with liquid carbon dioxide. Even in modern quality control laboratories you will find a number of tradi- tional methods for the identification of single flavour compounds, for example the estimation of optical rotation, refractive index, density and melting point, since these methods are generally accepted, effective and less time-consuming. Especially for the purpose of fast identification checks of more complex sys- tems, spectroscopic methods, above all infrared (IR) and near-IR spectroscopy, are gaining more and more importance. Numerous analyses in the quality control of most kinds of samples occur- ring in the flavour industry are done by different chromatographic procedures, for example gas chromatography (GC), high-pressure liquid chromatography (HPLC) and capillary electrophoresis (CE). Besides the different IR methods mentioned already, further spectroscopic techniques are used, for example nuclear magnetic resonance, ultraviolet spectroscopy, mass spectroscopy (MS) and atomic absorption spectroscopy. In addition, also in quality control mod- ern coupled techniques like GC-MS, GC–Fourier transform IR spectroscopy, HPLC-MS and CE-MS are gaining more and more importance.

14.6 Specific Safety Aspects 307 14.5 Sensory Evaluation Over the last few decades scientist have developed sensory testing from the ear- liest individual examinations into a formalised, structured and codified meth- odology. Subsequently, sensory tests have become valuable, important and pre- cise tools in quality control, which are equivalent to the physical and chemical methods used. However, sensory testing is not only a tool in quality assurance, but also in grading, product development and marketing, as well as for the cor- relation between specific chemical/physical properties of a food and the effect on the human sensorial perception. Besides smell and taste, the sensorial evaluation of raw materials and final products covers trigeminal impressions (e.g. hot) and visual impressions like colour, opacity and particle size. In order to obtain reproducible results, special care must be given to panel selection and panel education, testing facilities, sample presentation and the design of each test. Modern sensory facilities display a kitchen/laboratory for sample preparation as well as separate sensory booths with controlled air and lighting. The evaluation of the sensory results can be supported by specialised computer software packages. The most frequently used tests in quality control in the flavour industry are paired-sample comparison tests, and triangle tests, which are often combined with the description of deviation from a reference item. For the selection and training of panellists, further test methods are used, for example ranking tests for colour, taste and odour, threshold detections (taste, off-flavour), colour blindness tests and odour identification tests [6]. 14.6 Specific Safety Aspects Over the last few decades, safety has become one of the most important topics related to food. From this view, quality control of vegetable raw materials has at first to cover the following issues: natural and anthropogenic contaminants (e.g. heavy metals, pollution from industrial and private combustions, not pro- fessionally deposited waste products, radionuclides), residues of fertilisers (e.g. nitrate), plant-conditioning and plant-protective agents, filth, pests, the micro- bial status and the occurrence of microbial toxins. It is not possible to discuss all these aspects in detail; however, with a focus on herbs and spices, two of them should be stressed more thoroughly. For further information, see [8]. As background, it has to be remembered that a large number of our spices are imported from less-developed tropical and subtropical countries which often lack the necessary consistency in the application of a quality-oriented cultiva- tion and adequate processing. The hot, humid climate prevailing in many culti- vation countries, the mostly simple, unpretentious production conditions, and

308 14 Industrial Quality Control the often inadequate instruction of farmers and farm hands give rise to funda- mental problems [8]. 14.7 Microbial Aspects and Microbiological Methods The fact that food-borne infections and intoxications increased during the last decade of the twentieth century in several European countries underlines the necessity of microbial investigations of food as well as their additives and ingre- dients [9]. However, compared with the physicochemical methods, microbio- logical tests play a less important role in the quality control of the flavour indus- try. Most of the liquid flavours contain solvents like ethanol, propylene glycol or edible oils in concentrations between 70 and 90% so that they possess bacterio- static or bactericidal properties. Moreover some of the aroma compounds pos- sess the same properties, so routine microbial investigations are not necessary for many of the raw materials and final products [6]. Microbiologically critical products of the flavour industry are above all emulsified, pasty or dry products as well as the agricultural raw materials of animal and plant origin. According to [10], for microorganisms each particular type of vegetable pro- vides a unique environment in terms of type, availability and concentration of substrate, buffering capacity, competing microorganisms and perhaps plant antagonists. So it is not surprising that every natural plant material harbours numerous and varied types of microorganisms, among which pathogenic or food-spoiling species like salmonella, staphylococci, bacilli, clostridia, entero- haemorrhagic Escherichia coli or moulds may occur. In the case of herbs and spices—one of the important sources for the flavour industry—the number and composition of the microflora is above all influenced by the part of the plant used for spice (leaves, seeds, flowers, etc.) and beyond that by harvesting, post- harvest treatment, the drying process as well as storage and transport condi- tions. In order to give an idea about the microbial load of herbs and spices Table 14.1 surveys the microbial numbers which can be found on untreated ground spices. Among the pathogenic microorganisms which might be available, special at- tention has to be paid to the Enterobacteriaceae. A salmonella outbreak in 1993 caused by paprika-powdered potato chips showed the importance of these mi- croorganisms even with spices [12]. Burow and Pudich [13] reported that 5.4% of 317 samples of paprika powder and 10.1% of 139 snack products investigated in 1993 and 1994 were salmonella-positive. The differentiation of the isolated salmonellae revealed a number of seldom-found serotypes, which led to the conclusion that the contamination had its origin in the producing countries. The evaluation of the aforementioned outbreak seems to be of practical con- sequence for quality control throughout the food industry. On the basis of the results of this evaluation and on the basis of the knowledge that there are strains of salmonella which owing to the existence of plasmids or prophages are able

14.7 Microbial Aspects and Microbiological Methods 309 Table 14.1 Microbial counts (cfu/g) of untreated ground spices [11] Spice Total germ count Coliforms Moulds Allspice 1×105–5×106 <102–1×103 5×102–5×104 Anise 2×105–2×106 <102–1×104 <102–1×103 Basil 2×104–4×105 <102 <102 Caraway 1×105–5×106 <102–1×103 <102–1×104 Cinnamon 2×103–5×104 <102 2×102–5×103 Cloves 5×104–1×107 <102 <102 Coriander 1×103–1×107 <102 <102–5×104 Fennel 5×104–1×105 <102 1×102 Garlic 5×104 <102–5×102 <102–5×102 Ginger 1×104–1×107 <102 <102 Laurel (bay) 1×103–5×104 <102 2×102–2×104 Mace 1×104–5×104 <102–1×103 1×102–5×104 Marjoram 2×105–1×106 5×102 5×103–5×104 Oregano 5×103–5×104 <102 5×102–5×103 Paprika 1×105–5×105 <102 2×102–5×102 Pepper black 5×105–1×107 1×102–5×104 5×102–2×104 Pepper white 1×104–5×105 <102–1×103 <102–1×105 Tarragon 5×104 <102–1×103 <102 Thyme 5×105–1×107 <102 5×102–1×104 Turmeric 1×104–2×107 <102–1×103 <102–3×103 to change their virulence and thereby their minimal infectious doses, it is as- sumed today that under special circumstances one single salmonella could be infectious. In this case, any finding of salmonellae in a food has to be regarded as a health risk. Consequently it was decided by the European authorities that salmonella should not be detectable in 25 g of herbs or spices. As a consequence, the importing and manufacturing companies have estab- lished comprehensive and efficient examination procedures for raw material ac- ceptance. The major problem in this relation is the fact that microorganisms as well as other contaminants normally are not homogenously distributed within the product. For that reason detailed sampling plans have been developed; the best known one is the Foster plan [14], which is specialised for salmonella de- tection. Industrial microbiological quality control normally covers the total count of germs, the counts of yeasts and moulds, coliform bacteria and E. coli. In special cases these investigations are complemented by the detection of Staphylococcus aureus, salmonella and listeria. The German Association for Hygiene and Microbiology has published mi- crobial approximate and warning values for spices which should be given to the consumer or which should be used in the production of foods which do not

310 14 Industrial Quality Control Table 14.2 Microbial approximate and warning values for spices which should be given to the con- sumer or which should be used in foods without any further germ reduction treatment [15] Organism Approximate value (cfu/g) Warning value (cfu/g) Salmonella – Not detectable in 25 g Staphylococcus aureus 1.0×102 1.0×103 Bacillus cereus 1.0×104 1.0×105 Escherichia coli 1.0×104 – Sulphite-reducing clostridia 1.0×104 1.0×105 Moulds 1.0×105 1.0×106 undergo further heat treatment. The data are summarised in Table 14.2 and can be used as a guideline for the microbiological assessment of spices. Once again, it can be seen that salmonella should not be detectable in 25 g of the spice. The traditional microbiological methods are very time consuming and some- times limited concerning their interpretation. For that reason fast analysis meth- ods as well as automated methods have been developed; the latter are often used in specialised microbiological laboratories. During the last few years more and more modern biotechnological methods have been implemented into quality control, for example the enzyme-linked immunosorbent assay or more recently the polymerase chain reaction, which allows the detection of very specific mi- croorganisms. The occurrence of moulds on or within vegetable raw materials represents a serious problem, since the topic is of increasing interest and is doubly prob- lematic: for one thing, in many countries mouldy foodstuffs are considered dis- gusting, regardless of the sanitary risk they present; secondly, the presence of moulds always implies the risk of mycotoxin formation. Besides their acute tox- icity, some of the mycotoxins have been found to be teratogenic, mutagenic and/ or carcinogenic. The most critical spices concerning the possible occurrence of mycotoxins are coriander, paprika, chillies and nutmeg, a special problem being that moulds growing within capsicum fruits as well as within nutmeg nuts can- not be detected from the outside. 14.8 Residues of Plant-Conditioning and Plant-Protective Agents For several reasons, residues of plant-conditioning and plant-protective agents represent a highly complex problem. So is it often not quite clear what to look

14.9 Biologically Active Substances 311 for since for different reasons sometimes substances are used which are not in- tended for the particular crop. This can occur when the intended preparations are not available or by crossing-over effects from intensively used plantations which may be located close to the often small spice parcels. Otherwise up to now there are no global methods for the detection of residues of plant-conditioning and plant-protective agents available, which is above all owing to the high num- ber of active substances as well as to their multifarious chemical structures. Another serious problem is the fact that the maximal accepted residues for the different plant-conditioning and plant-protective agents are not harmonised and that, for example, in Germany for some substances the maximum value for residues in herbs and spices is generally set at 0.01 ppm. This value was solely set by a political decision and has no proven toxicological background. Moreover in a number of cases this value is near the lowest detection value of the particular substance. 14.9 Biologically Active Substances One critical subject concerning the quality of herbs and spices as well as of flavours is the discussion of so-called biologically active substances. Among the estimated 60,000–100,000 different secondary plant metabolites numerous substances can be found which have considerable effects on the human organism, i.e. they are biologically active. This is not astonishing, since most of the vegetables used as herbs and spices today were originally used for their pharmaceutical properties. Since Paracelsus (1493–1541) we know that it is only a question of dose whether something is poisonous or not, and so it is not surprising that among the secondary plant metabolites substances can be found which have to be seen as critical from a toxicological point of view. Often we have to ask the question how to reconcile the obvious contradiction that animal experiments with isolated substances indicate their carcinogenic potential, whereas daily experience obviously does not indicate a serious health risk for humans. Moreover, a recently undertaken comparison of the genotoxicity of estragol with that of tarragon (the plant which contains the highest percentage of estragol) showed that untreated tarragon has a genotoxic potential under the conditions of the test carried out. However, this activity is clearly lower than that observed with methylchavicol at the same concentrations. Dried tarragon, on the other hand, showed no genotoxic potential under the same experimental conditions (study conducted for the European Spice Association, 2005, un- published). Since the idea of a zero risk of food in principle cannot be realised, it is not least an ethical request to set the assessment of naturally occurring biological active substances on a new basis corresponding to the rules of a scientifically substantiated risk assessment.

312 14 Industrial Quality Control References 1. Dürrschmid K, Zenz H (2000) Ernährung/Nutrition 24:119 2. Acree TE, Teranishi R (1993) Flavor Science—Sensible Principles and Techniques. ACS Pro- fessional Reference Book. American Chemical Society, Washington 3. Blech ZY (2006) Kosher Food Production. Blackwell, Oxford 4. Riaz MN, Chaudry MM (2004) Halal Food Production. CRC, Boca Raton 5. Matheis G (1999) Quality Control of Flavourings and Their Raw Materials. In: Ashurst R (ed) Food Flavorings, 3rd edn. Aspen, Gaithersburg, p 153 6. Lösing G, Matheis G, Romberg H, Schmidt V (1997) Dragoco Rep. 42:93 7. Bauer K, Garbe D, Surburg H (1997) Common Fragrance and Flavor Materials. Wiley-VCH, Weinheim 8. Buckenhueskes HJ (1998) Z. Arznei- Gewürzpflanz. 3:21 9. Schmidt K, Kolb H (1996) Verbraucherdienst 41:4 10. Daeschel MA, Andersson RE, Fleming HP (1987) FEMS Microbiol. Rev. 46:357 11. Buckenhueskes HJ, Rendlen M (2004) Food Sci. Biotechnol. 13:262 12. Lehmacher A, Bockemuehl J, Aleksic S (1995) Epidemiol. Infect. 115:501 13. Burow H, Pudich U (1996) Fleischwirtschaft 76:640 14. Foster EM (1971) J. AOAC 54:259 15. DGHM—Deutsche Gesellschaft für Hygiene und Mikrobiologie (1988) Dtsch. Lebensm.- Rundsch. 88:78

15 Advanced Instrumental Analysis and Electronic Noses Hubert Kollmannsberger, Siegfried Nitz Lehrstuhl für Chemisch Technische Analyse, TU München, Weihenstephaner Steig 23, 85350 Freising, Germany Imre Blank Science Department, Nestlé Product Technology Center, 1350 Orbe, Switzerland 15.1 Introduction Since the development of suitable gas chromatographic methods in the 1960s, researchers have been able to identify thousands of volatile compounds in foods, essential oils and fragrances. High-resolution gas chromatography (GC) combined with mass spectrometry (MS) became the key technique used for quantification and identification of flavour compounds. Individual flavours were found to be complex mixtures containing hundreds of compounds in con- centrations ranging from percent to trace (nanograms per kilogram) levels. Ad- ditional information is obtained by smelling the GC effluent after its separation. This technique is called GC-sniffing or GC-olfactometry (GC-O). To determine the relative sensorial importance of a volatile compound to the overall flavour, odour activity values are calculated by relating the measured concentration of a compound to its odour threshold. Since this does not work with unidenti- fied and trace compounds, other methods using GC-O for the determination of the specific contribution of gas chromatographically separated compounds have emerged: aroma extract dilution analysis (AEDA), combined hedonic aroma re- sponse measurements (CHARM) and OSME (a time-intensity rating method) [1, 2]. These sniffing techniques clearly demonstrated that of all these volatile components only a few contribute to the characteristic odour and only in some cases “character-impact compounds” could be found. Therefore, the compre- hensive identification of all substances is no longer the main goal. The aim of modern flavour research is now focused on those volatile constituents which ei- ther independently or in combination produce a characteristic aroma response. This forced flavour chemists to develop new advanced techniques, both for iso- lation and concentration as well as for separation and identification. Besides classical headspace analysis, simultaneous distillation–extraction and solvent extraction, new sampling and enrichment developments include solvent-assisted flavour evaporation (SAFE) [3] and sorptive techniques like SPME solid-phase microextraction (SPME) [4,5] and stir-bar sorptive extrac- tion (SBSE) [6], which are treated in a dedicated chapter in this book. This contribution will deal with advanced developments of GC techniques for im- provement of separation and identification (classical multidimensional GC, or

314 15 Advanced Instrumental Analysis and Electronic Noses MDGC, and comprehensive two-dimensional GC, or GC×GC), faster separa- tion techniques (fast GC), fast methods for quality assessment or process con- trol in the flavour area (“electronic noses” and fingerprinting MS) and on-line time-resolved methods for analysis of volatile organic compounds (VOCs) such as proton-transfer reaction MS (PTR-MS) and resonance-enhanced multi-pho- ton ionisation coupled with time-of-flight MS (REMPI-TOFMS). The scope of this contribution does not allow for lengthy discussions on all available tech- niques; therefore, only a selection of developments will be described. 15.2 Multidimensional Gas Chromatography Single-column gas chromatographic analysis has become the standard approach for separation of volatile and semivolatile constituents in numerous applications; however, this does not necessarily provide the best analytical result in terms of unique separation and identification. There is considerable opportunity for peak overlaps, both on a statistical basis and also on the basis of observed separations achieved for real samples [7–9]. 15.2.1 Classical Multidimensional Gas Chromatography In order to expand the analytical separation, chromatographers have developed a range of solutions based on more than one separation space. Termed MDGC, it consists of an arrangement of two or more columns where distinctive seg- ments of effluent from the first column are fed into one or more secondary col- umns (Fig. 15.1). The entire procedure is enabled by the presence of a specific transfer device between the two columns. Cortes [10] and Bertsch [11, 12] have presented a comprehensive discussion of conventional MDGC technology and their contri- butions are recommended to readers who wish to have a more detailed outline of MDGC and its applications. In conventional two-dimensional GC (Fig. 15.1a), discrete fractions of ef- fluent (heart cut) are diverted into a secondary column, generally with a dif- ferent polarity. This arrangement has the disadvantage that components from different cuts may intermingle in the second column and thus can no longer be correlated. Application of parallel traps is one possibility (Fig. 15.1b) to solve this problem. Of course, this is achieved at the expense of increasing the total analysis time, since each fraction delivered to the trap must be individually ana- lysed in a conventional GC procedure. This disadvantage can be circumvented by using the highly complex arrangement shown in Fig. 15.1c, where each cut is directed into a separate column. The mechanism by which effluent is switched from the first to the second column is critical. Column flows are diverted basically by using valves or valve-

15.2 Multidimensional Gas Chromatography 315 Fig. 15.1 Basic arrangements in multidimensional gas chromatography (MDGC): a conventional, b multitrap, c multicolumn less Deans-type switching devices. Accurate descriptions of the most important MDGC interfaces have been reported in the literature [12]. A truly versatile ar- rangement is presented in Fig. 15.2 by way of example. It consists of a dual oven with a valveless Deans-type interface (Live-T switch- ing device) between the columns, an intermediate effluent monitor detector and facilities for flow reversal. Additionally, a device (total transfer) to focus and reinject the desired effluent fraction after the first column allows the combi- nation of a high-capacity precolumn (packed or thick-film wide bore) with a high-resolution capillary or chiral column. Trapping a heart cut before intro- duction into the second dimension makes the second separation independent of the first. Depending on the application, the instrument incorporates inlets for liquid injection, dynamic headspace enrichment, thermal desorption and facili- ties for SPME, SBSE and headspace sorptive extraction (HSSE). After suitable enrichment and initial separation on a high-capacity column, appropriate heart cuts are transferred to the main column. After final separation, the substances of interest can be directed via a second Live switching device to an MS, isotope

316 15 Advanced Instrumental Analysis and Electronic Noses Fig. 15.2 A two-dimensional GC (GC×GC) system. D1 intermediate effluent monitor detector, column 1 high-capacity precolumn, column 2 high-resolution capillary or chiral column ratio MS (IRMS), 14C or Fourier transform IR detector or into individual traps. Accumulation of sufficient material (preparative scale isolation) for further characterisation by off-line instruments such as an NMR spectrometer can be achieved by collection of selected heart cuts from the second column in separate traps after a certain number of GC enrichment cycles [13,14]. Apart from petrochemical and environmental applications [12], classical MDGC was/is used in the flavour field mainly for enrichment and identification of odorous trace compounds in complex mixtures, or for authenticity evalu- ation by chiral separation or isotopic ratio determination. In Table 15.1 some typical applications are given. In order to illustrate the potential of MDGC when dealing with complex mixtures, an application to determine off-flavour compounds in defective coffee beans is given by way of example (Fig. 15.3) As can bee seen, MDGC is a targeted analysis applicable to a specific ap- plication. Heart cutting can be effectively applied to a small number of regions of interest in the chromatogram. Transferring them to a second column gives enhanced resolution of the heart-cut zones owing to the different column selec- tivity. Care should be taken that compounds from one cut do not interfere with the separation of another cut. With respect to the maximum numbers of cuts achievable in a single MDGC run, this is dependent on the sample type and on the analytical conditions. A practical implementation of a large number of heart cuts in order to completely separate a complex mixture in a single run is just not possible. Nevertheless, heart-cutting MDGC may be considered for target ap- plications the most appropriate analytical choice.

15.2 Multidimensional Gas Chromatography 317 Fig. 15.3 Dynamic headspace MDGC analysis of three coffee beans. a Chromatogram of precol- umn showing heart cuts at retention of peasy-like odours detected by sniffing the eluent from the precolumn. b Heart cut from normal beans on the main column. c Heart cut from defective beans on the main column [14] If the speed of the secondary separation is high enough to separate a cut from the first separation while the next cut is being collected, the complete two-di- mensional chromatogram could be constructed. A new type of instrumentation was developed to accomplish this goal. The technique, called comprehensive GC×GC, was introduced in 1991 by Liu and Phillips [30]. It expands the MDGC method into a generally usable format that does not rely on targeting specific zones of a first-dimension analysis. 15.2.2 Comprehensive Two-Dimensional Gas Chromatography Comprehensive GC×GC uses the whole two-dimensional separation space to generate resolution, provided that the individual dimensions are orthogo- nal. GC×GC consists of two chromatography columns, serially coupled, with a modulation mechanism at their junction. A typical column set is composed of a standard low-polarity column (first dimension), with typical dimensions of 25 m × 0.25-mm internal column diameter × 0.25-μm film thickness, coupled to a much shorter and more polar second column (second dimension) (or a column providing a separation mechanism capable of further differentiating target sample components), with dimensions of 1 m × 0.1-mm internal column

318 15 Advanced Instrumental Analysis and Electronic Noses Table 15.1 Selected applications of multidimensional gas chromatography in flavour research Enrichment of trace com- Application References pounds from interfer- [15, 16] ing complex mixtures Identification of odour- active undecaenes in [14, 17, 18] Selective transfer of compounds fruits and vegetables [19] of interest into a second column [14] Identification of a peasy [20–24] Preparative-scale enrichment off-flavour in coffee beans [24–26] Identification of musty/earthy off-flavours in wheat grains [27] [23, 24, 28] Identification of a rotten off-flavour in car mats [13, 29] Enantioresolution of lac- tones for authenticity con- trol of fruit products Enantioresolution of ter- penes for authenticity control in essential oils Enantioresolution of nor-carotenoids 13C/12C isotope ratio de- termination for authentic- ity control of flavours Isolation of terpenes from essential oils diameter × 0.1-μm film thickness. The modulator interfaces the two coupled columns, and is responsible for the quantitative transfer and compression of all solutes, or a representative fraction thereof, from column 1 to column 2. To ac- cumulate the analyte in narrow bands in the modulator, either elevated tem- perature (the thermal sweeper) to accelerate the solute into a narrow band or cryogenic means to retard the analyte and cause on-column trapping is used [31]. The operation of a dual-jet cryogenic system is explained in Fig. 15.4. The resulting very sharp peaks are then released onto the short fast column 2. The modulator actually collects eluent from column 1 every few seconds (gen- erally 2–9 s), and so an individual chromatographic peak is actually sliced into many fragments. Figure 15.5 demonstrates how two overlapping peaks are ef- fectively deconvoluted into two interleaved series of pulses. Each fragment is focussed and pulsed to column 2 for fast analysis. Because modulation is a mass-conservative process, the peak height increases to ac- commodate the reduction in peak width; thus, greater analytical sensitivity is obtained. Provided that column 2 can resolve the substances focussed by the

15.2 Multidimensional Gas Chromatography 319 Fig. 15.4 A dual-jet cryogenic modulator. a Right-hand-side jet traps analytes eluted from the first column; b right-hand-side jet switched off, cold spot heats up rapidly and analyte pulse is released into the second column; simultaneously, left-hand-side jet switched on to prevent leakage of first- column material; c next modulation cycle is started (adapted from [32]) Fig. 15.5 Illustration of how two overlapping peaks δ and φ emerging from the first column (a) are resolved in GC×GC after passage to the second column (b) [31]. Reprinted from Marriott, P., Shellie, R., Principles and applications of comprehensive two-dimensional gas chromatography. Trends Anal. Chem. (2002), 21:573–583 with permission from Elsevier

320 15 Advanced Instrumental Analysis and Electronic Noses modulator, compounds which would coelute under conventional single-column GC (Fig. 15.5a) will be separated because of the modulation process, and differ- ent pulses can be assigned to different compounds depending upon their reten- tion times (Fig. 15.5b). The data are generally presented in a two-dimensional plane, which plots the retention time on column 1 (minutes) against the retention time on column 2 (seconds), or a three-dimensional plot where the detector response is also in- cluded. In performing a carefully tuned GC×GC experiment, the peak capacity of the overall separation is approximately equal to the product of the peak ca- pacities of the individual separation steps [33]. Thus, the opportunity to char- acterise mixtures fully is far greater using GC×GC than for both single-column GC and MDGC. Regardless of the type of analysis, many variables need to be adjusted for op- timal performance. It is generally desirable that chromatography in the second column is complete before another aliquot is transferred from the primary col- umn. The need for rapid elution from the second column sets practical limits. In order to get maximum separation performance, each individual first-dimension peak should be modulated into several fractions (in general some five to ten). For that purpose, the ratio of separation speeds between the second and first dimen- sions must be at least on the order of 50 [30]. Generally, complete elution within a time frame of 2–8 s is required for column 2. With a half peak width of 0.2 s in the second dimension, and the acceptance that at least ten points per peak half width are required to be suitably measured by a chromatographic detector, fast electronics for detection and data collection are needed. In GC×GC to date, detection techniques employed include flame ionisation detection (FID), sulphur chemiluminescence detection (SCD), atomic emission detection (AED), electron capture detection (ECD), nitrogen chemiluminescence detection (NCD) and MS detection (both time-of-flight MS, or TOFMS, and quadrupole MS, or qMS) [34]. In order to conduct GC×GC, the scan speed of the detector is critical: each detec- tor used must be critically evaluated with respect to operational considerations that may limit or affect performance. A possible alternative detection technology, described recently for fast GC, is the surface acoustic wave (SAW) sensor [35]. Several approaches are reported to perform peak quantification in GC×GC. The most common one integrates all individual second-dimension peaks by means of conventional integration algorithms, and then sums all peak areas be- longing to one compound. For another method, firstly a so-called base plane is subtracted, and subsequently three-dimensional peak volumes are calculated by means of imaging procedures. Although the peak capacity of GC×GC is high, peak overlapping in two-dimensional separation is very possible, especially for highly complex samples. Chemometric methods, like the generalised rank an- nihilation method (GRAM), have been used to resolve and quantify severely overlapped GC×GC peaks. Some other methods have also been used, like curve fitting, wavelet analysis [34] and orthogonal projection resolution [36]. The practicability and potential of comprehensive GC×GC coupled to TOFMS (GC×GC-TOFMS) for the analysis of complex mixtures is illustrated in the fol- lowing application [37].

15.2 Multidimensional Gas Chromatography 321 Figure 15.6 shows the separation achieved for the essential oil of Coriandrum sativum using GC×GC. The identity of the compound was elucidated and con- firmed primarily from the MS library matches as well as by comparing the first- dimension retention index with reference libraries. GC×GC-TOFMS allowed the identification of 81 compounds, compared with only 41 compounds identi- fied by conventional GC-qMS. A great advantage of GC×GC is that homologous series of compounds form linear relationships in the two-dimensional separation plane. Figure 15.7 shows the homologous series of compounds identified in C. sativum essential oil. This provides another method to confirm compound identity and allows easy discrimination between series of isomers. For example, (E)-2-alkenals and (Z)-2-alkenals exhibit very similar mass spectra but are easily distinguished by GC×GC (Fig. 15.7). In addition, many of the heavier compounds were not pres- ent in the mass spectral libraries (and so were often misidentified as lower-mass homologues) but were easily identified using their homologous series. It is important to appreciate that whilst the GC×GC analysis might not be any faster overall than normal capillary GC, within a similar analysis time, higher sensitivity, greater peak resolution (and hence one could expect greater preci- sion of analysis) and a fingerprint pattern that may contain much subtle infor- mation on the chemical class composition of samples, which cannot be achieved in any other way, is obtained. Selected examples of application areas of GC×GC are presented in Table 15.2. For further details about instrumentation and applications of comprehensive GC×GC, the contributions in [32, 38, 39] should be consulted. Fig. 15.6 GC×GC time-of-flight mass spectrometry (TOFMS) total ion chromatogram of corian- der essential oil as a contour plot. Reprinted with permission from [37]. Copyright (2005)

322 15 Advanced Instrumental Analysis and Electronic Noses Fig. 15.7 Apex plot showing the homologous series of compounds present in coriander essential oil. The apex of each peak is plotted as the second -dimension retention time against the first-di- mension retention time. Reprinted with permission from [37]. Copyright (2005) Wiley 15.3 Fast Gas Chromatography Conventional GC allows effective separations of complex natural mixtures, but this is frequently achieved at a high cost in time. This becomes a limiting factor, especially for laboratories with a high sample throughput and/or where the need for quick results for the determination of quality and authenticity are required. In recent years there has been increasing interest, within the chromatographic community, towards the development of faster separation methods without con- siderable loss of resolving power. Various approaches have been theorised and developed with various proposals: shorter column lengths [66], reduced inter- nal column diameter and stationary phase thickness (narrow-bore column) [67, 68], microparticle-packed capillary columns [69, 70], multicapillary columns [71], vacuum-outlet conditions [72], turbulent flow [73] and helically coiled columns [74]. Reviews describing the most important existing high-speed GC methods, also trade-offs and compromises in terms of sensitivity and/or selec- tivity in combination with MS, have been published [75, 76]. The narrow-bore column approach is a very effective and is the most popular way of increasing analysis speed. Substantial reductions in analysis times are achieved by exploiting two factors: a shorter column length and the application of higher than optimum average linear velocities. Operating under optimum

15.3 Fast Gas Chromatography 323 Table 15.2 Selected applications of comprehensive two-dimensional gas chromatography Plant constituents Enantiomeric alkaloids [40] Volatiles from germander [41] Food Tobacco essential oil [44] Lipids in lanolin [50] Fragrances Coriander leaves essential oil [37] Human breath Pistacia vera essential oil [58] Petrochemicals Sandalwood oil [55] Environmental Tea tree and lavender essential oils [60] Origanum micranthum essential oil [61] Hop essential oil [65] Flavour compounds in butter [48] Volatiles in strawberry cultivars [54] Roasted coffee bean volatiles [56] Methoxypyrazines in wine [51] PCBs in milk and cheese [53] Trace odorants in sour cream [63] Fatty acid composition in foods [62] Citrus essential oil [64] Yeast cell metabolites [45] Perfume analysis [52] Allergens in fragrances [59] Volatile organic compounds [42] Oil spill [43] Composition [46] Diesel fuel hydrocarbons [47] Sulphur compounds in crude oil [57] Polychlorinated alkanes in dust [49] experimental conditions, a 10 m × 0.1-mm internal diameter, 0.1-µm film thick- ness column is characterised approximately by the same resolving power as a 25 m × 0.25-mm internal diameter, 0.25-µm film thickness column [67]. Figure 15.8 shows the chromatograms of conventional GC and fast GC analysis of a lemon oil with the aforementioned columns [77].

324 15 Advanced Instrumental Analysis and Electronic Noses As can bee seen, 57 components were separated with both methods. The fast GC technique performs the same separation within 9 min, a speed gain of a factor of 5 compared with the conventional method. A lime oil sample, in a ap- plication aimed at quality control, was separated satisfactorily in only 90 s on a 5 m × 0.5-mm internal diameter, 0.05-µm film thickness column [78]. Other applications include essential oil analyses [79,80], flavour volatiles in fruits [81], fatty acid composition [82] and pesticides [83]. Fast GC requires instrumentation provided with high split ratio injection sys- tems because of low sample column capacities, increased inlet pressures, rapid oven heating and fast electronics for detection and data collection. Hydrogen is generally used as a carrier gas because of the flatness of its Van Deemter height equivalent to the theoretical plate (HETP)–µ curve, which allows higher linear gas velocities to be applied than the optimum without substantial loss of reso- lution. Shorter columns, thinner films and smaller internal diameter columns used in fast GC require smaller amounts of sample to be injected to prevent overloading of the column. This in turn causes the detection limits to be higher in fast GC. This is a problem when working with trace levels of analytes. It has been shown that using fast temperature programming is a better way than using faster flow rates to decrease the analysis times [84]. This parameter has been ignored in many studies, but it offers valuable time savings with some added benefits. Shorter columns with typical internal diameters (e.g. 0.25 mm) and film thicknesses can be used, without much loss in sample capacity. It is important to ensure that the data collection rate is fast enough for peaks with low retention times in order to ensure good reproducibility of all peak parameters. For modern instrumentation, this is generally not a problem; for example, FID detectors are typically able to achieve a data acquisition rate of 50–250 Hz using the standard instrument configuration. The data sampling rate of a typical quadrupole benchtop mass spectrometer is not always fast enough for very fast GC analyses, especially for quantitation. Typical spectral acquisition rates of scanning mass spectrometers, such as the ion trap, the quadrupole and the sector instruments, are limited to a maximum of ten to 20 spectra per second in the full-scan mode. This is just on the edge of applicability for fast GC. If faster detection is required, non-scanning TOF analysers are an alternative. TOFMS can provide up to 500 full spectra per sec- ond and allow accurate detection of peaks with peak widths in the millisecond range (very fast GC), while still providing high-quality spectra [85]. In terms of ultimate potential instrument performance, sensitivity is sacrificed for gains in speed in TOF [76]. In the literature, much of the discussion about fast GC-MS originates from the chromatographer’s point of view, and a chromatographer tends to prefer baseline resolution between peaks. Although more selectivity in the separation can be beneficial in some respects, in other respects the time spent to resolve coeluting compounds by GC is wasted if the compounds can be adequately resolved by the MS detector. Mass-spectral deconvolution software is an effective and efficient tool to resolve coeluting peaks in GC-MS and thus is very important for fast GC-MS. Humans simply cannot conduct adequate back-

15.3 Fast Gas Chromatography 325 ground subtraction in a complex chromatogram, and a highly trained person could spend hours trying to do what an adequate deconvolution program can do in seconds. The future usefulness of fast GC-MS depends to some extent upon improvement of existing approaches and commercialisation of interesting Fig. 15.8 Conventional GC (a) and fast GC (b) chromatograms of a lemon essential oil [77]. Copy- right (2003) American Chemical Society

326 15 Advanced Instrumental Analysis and Electronic Noses new techniques. Moreover, a greater emphasis is needed to rationalise overall laboratory operations and sample preparation procedures if fast GC-MS is to become implemented in routine applications [76]. 15.4 Electronic Noses In the food industry there exists an increased demand for fast, simple and sen- sitive methods of assessing volatiles, for identification, authentication, process control, and product blending or formulation. Since the quality of raw materials and processed products is determined frequently by the volatiles characteristic for a particular odour, objective methods for aroma and flavour evaluation are needed. By this means contaminations or off-flavours of such products could also be detected. Actually for this purpose human sensory panels of trained experts are used in the food and aroma industry, complemented by more ob- jective, but time-consuming analytic methods, such as GC. In order to be able to interpret the laboratory analysis, those GC methods often have to be com- bined with MS or simultaneously used “sniffing” lines. However, even the most sophisticated analytic methods cannot fully replace the human nose, as only our olfactory sense can determine whether a compound is relevant to a specific odour. Therefore, it is not surprising that repeated efforts have been made over the years to introduce instruments operating on a similar principle to the hu- man nose. The instruments comprise an array of electronic chemical sensors with partial specificity and an appropriate pattern-recognition system, capable of recognising simple or complex volatile mixtures. These systems would in most cases not replace but would complement conventional analyses of vola- tile compounds by sensory methods and by classic analytical techniques. The arrangements realised are often called “electronic noses” or “artificial noses” although the only aspect in common with our odour-sensing organ is the func- tion. The operating principle, the number of sensors as well as the sensitivity and selectivity are, however, very different. This is why they should better be called “multisensor array technology”. Within the last few years, gas sensors together with associated pattern-recog- nition techniques have been used to differentiate and identify complex mixtures of volatile compounds. Potential areas of application include the food indus- try [86], perfumery, the chemical industry, the pharmaceutical industry, the tobacco industry [87], cosmetics [88], health and environmental control [89], where they are intended to be used for quality control of raw and manufactured products, monitoring of freshness and maturity, evaluating process effective- ness, prediction of shelf life, microbial pathogen detection [90, 91] and authen- ticity profiling. The major categories involved in the development of electronic noses and the principles of sensor design and technology will be described in the following section.

15.4 Electronic Noses 327 15.4.1 Catalytic or Metal Oxide Sensor A catalytic or metal oxide sensor (MOS) consists of an electrically heated (250– 450 °C) ceramic pellet upon which a thin film of tin(II) oxide doped with pre- cious metals is deposited [92]. Tin(II) oxide is an n-type semiconductor and when oxygen adsorbs on the surface, one of the negatively charged oxygen species is generated, depending on the temperature. This results in the surface potential becoming increasingly negative and the electron donors within the material be- come positively charged. When an oxidisable material comes into contact with the sensor surfaces, the adsorbed oxygen is consumed in the resulting chemical reaction. This reduces the surface potential and increases the conductivity of the film. As a result the electrical resistance of the sensors will change. Disadvanta- geous are the relatively poor selectivity, which can be to some extent improved by dopants and temperature adjustment during the measurement, sensor drift over time, and finally the high power-consuming operation temperature. 15.4.2 Metal Oxide Semiconductor Field-Effect Transistor The metal oxide semiconductor field-effect transistor (MOSFET) sensor device is based on a field-effect transistor with a catalytic metal as the gate contact. The gate voltage controls the current through the MOSFET device. The gas mol- ecules will affect the voltage to the gate contact and thus change the current through the transistor. In a field-effect sensor, the interaction of gases with the catalytic gate metal induces dipoles or charges, which give an additional voltage to the gate contact. The choice of operation temperature, type of catalytic metal and structure of the metal influence the chemical reactions on the gate of the sensor, and thus the selectivity and sensitivity of the sensor. 15.4.3 Conducting Polymer Sensors Polymer materials like polypyrrole and polyaniline are conducting (or semicon- ducting) and show a variation in conductivity with sorption of different gases and vapours. The sensor response is not necessarily a linear relationship be- tween the analyte concentration and conductivity. Owing to their molecular structure, they show good sensitivities to polar compounds. The sensors display rapid adsorption and desorption at room temperature and specificity can be achieved by incorporating different metal ions in the structure of the polymer. Disadvantages include the reproducibility of fabrication (poor batch-to-batch reproducibility), strong humidity interference and the baseline drift over time

328 15 Advanced Instrumental Analysis and Electronic Noses owing to oxidation processes or changes in the conformation owing to exposure to inappropriate compounds. 15.4.4 Acoustic Wave Sensors One of the first sensors to be introduced was the thickness-shear mode (TSM) sensor, which, if the substrate is quartz, may commonly be termed the quartz crystal microbalance (QMB) or bulk acoustic wave (BAW) sensor. The sensor consists of overlapping metal electrodes at the top and bottom. This type can be used with up to 10-MHz fundamental resonance frequency with a standing resonant wave being generated where the wavelengths are related to the thick- ness. As the thickness increases (e.g. owing to added mass by deposition on the surface), the wavelength increases and the frequency decreases. Mass-sensitive devices, such as QMB or SAW oscillators, can detect a change of mass accurately via resonance frequency shifts. In contrast to other sensor technologies, these kinds of transducers generate a fully digital electrical output signal with all its advantages to further signal processing (e.g. no analogue to digital conversion, fewer electromagnetic compatibility problems). In a SAW device, an acoustic wave propagates along the surface, whereas in a QMB crystal the acoustic wave propagates in the bulk. Compared with the SAW devices, the resonance fre- quency of QMB sensors is an order of magnitude lower, allowing a wider tol- erance for temperature control of the oscillator electronics. When coated with gas-sensitive layers, both devices can detect gases or vapours. In contrast to SAW sensors, which use ultrathin layers to avoid damping of the oscillation, QMB sensors can also be coated with bulky layers. These coatings should have a high vapour permeability. SAW sensors can also be operated in the liquid phase and are theoretically more sensitive owing to their higher resonance frequency. In practice this advantage can often be compensated by using thicker coatings on the QMB devices. A large variety of chemical-sensitive materials can be de- posited onto mass-sensitive devices. The long-term characteristics of these de- vices are mainly dependent on the ageing and/or bleeding of coating materials. Stationary phases used in GC columns (e.g. silicone, carbowax) have been opti- mised in respect to these features over the last 20 years. A QMB sensor, coated with these polymers, shows reproducible and stable behaviour for a wide range of chemical substances. The abilities and limitations of this type of sensor have been described in detail recently [93]. 15.4.5 Mass Spectrometry Based Systems While most of the commercially available gas sensors are based on one of the aforementioned four major sensor technologies, MS sensors are based on a measuring technique well known and widely used for almost 30 years: MS.

15.4 Electronic Noses 329 Volatile sample components are introduced into the mass spectrometer with- out separation, thus creating a mass spectrometric pattern of fragment ions that describes the mixture of volatiles in the headspace [94–96]. Each fragment ion represents a potential sensing element and the intensity of the fragment ion is equivalent to the sensor signal. Theoretically, when performing a full-scan mea- surement from, for example, m/z 50 to m/z 300, one can choose up to 251 sen- sors to form a sensor array. However, it is not useful to work with such a great number of sensors, as in most cases only a very small number of fragment ions is needed for setting up a sensor array. The choice of fragment ions that should be selected to build up an array is based on knowledge and has to be correlated with the sample properties to be determined. Ion mobility spectrometry (IMS), which has the ability to separate ionic spe- cies at atmospheric pressure, is another technique that is useful for detect and characterising organic vapours in air [97]. This involves the ionisation of mol- ecules and their subsequent drift through an electric field. Analysis is based on analyte separations resulting from ionic mobilities rather than ionic masses. A major advantage of operation at atmospheric pressure is that it is possible to have smaller analytical units, lower power requirements, lighter weight and easier use. Other MS-fingerprinting techniques that are in commercial development are based on atmospheric pressure ionisation (API), resonance-enhanced multi- photon ionisation (REMPI) TOF and proton-transfer reaction (PTR). They are rapid, sensitive and specific and allow measurements in real time and may play an increasingly important role in the future development of electronic noses and tongues. 15.4.6 Other Sensor Technologies Apart from the aforementioned most frequently used sensor technologies, also selective electrochemical sensor combinations have been commercialised for use in dedicated applications. The combination of electrochemical CO, H2S, SO2 and NH3 sensors was used for quality and freshness control of foods like fish [98] and meat [99]. Combinations of MOSs and MOSFETs supplemented with a selective IR absorption sensor for carbon dioxide and a humidity sensor for measuring relative humidity were also described [100]. One of the technologies which may become relevant for gas sensing is bun- dled fibre optics, through which fluorescence is measured from photodeposited polymer-sensing elements. On one end of a fibre optic bundle, as many as 30 small regions of polymer fluorescent dye mixtures are photodeposited. A flash of light at an excitation wavelength is applied to the other end of the fibre optic, and fluorescence intensity at selected wavelengths from the polymer/dye mix is subsequently measured back through the fibre optic. Different polymer/dye combinations interact with gases differently, such that upon exposure to a given sample, the different regions or sensors provide unique information [101].

330 15 Advanced Instrumental Analysis and Electronic Noses Advances in the production, immobilisation and characterisation of mam- malian olfactory receptors led to the development of biosensors where isolated olfactory binding proteins were deposited on the surface of QMBs [102, 103] or were connected to nanoelectrodes [104]. Although still at the development stage, such an array-type device coated with different olfactory receptors will be a powerful and useful tool for detecting and discriminating odorants in the future. 15.4.7 Data Processing A generalised structure of an electronic nose is shown in Fig. 15.9. The sen- sor array may be QMB, conducting polymer, MOS or MS-based sensors. The data generated by each sensor are processed by a pattern-recognition algorithm and the results are then analysed. The ability to characterise complex mixtures without the need to identify and quantify individual components is one of the main advantages of such an approach. The pattern-recognition methods may be divided into non-supervised (e.g. principal component analysis, PCA) and su- pervised (artificial neural network, ANN) methods; also a combination of both can be used. PCA reduces multidimensional, partly correlated data, to two or three di- mensions. Projections are chosen so that the maximum amount of information is retained in the smallest number of dimensions. This technique allows the similarities and differences between objects and samples to be better assessed [105]. A neural network is a program that processes data like (a part of) the nervous system. Neural networks are especially useful for classification problems and for function approximation problems which are tolerant of some imprecision, which have lots of training data available, but to which hard and fast rules (such as laws of nature) cannot easily be applied. Fig. 15.9 Generalised structure of an “electronic nose”

15.4 Electronic Noses 331 Neural networks are trained with complete data sets consisting of input and output data. Typically one starts with a random configuration and calculates output data from the given input data. One compares the calculated output data with the output data of the complete data set and tries to minimise the error of the output data by varying the parameters of the network When the network has learned the complete data sets, one takes an independent collection of complete data sets to test the generalisation capability of the network. Both collections of complete data sets must be large enough and correctly distributed within the range of possible data. Perhaps the greatest advantage of ANNs is their ability to be used as an arbitrary function approximation mechanism which learns from observed data. However, using them is not so straightforward and a relatively good understanding of the underlying theory is essential. The danger is that the network overfits the training data and fails to capture the true statistical process generating the data, resulting in worse predicting ability [106]. 15.4.8 Applications, Potential and Limitations Most publications deal with the application to foods (Table 15.3), but published studies are also available covering other products, such as tobacco, cosmetics, health diagnostics and the environment [86]. The feasibility and limitations of using multisensor array systems in food and aroma applications will be discussed with an application intended to discrimi- nate hop varieties [149] by way of example The sensor responses generated in a measurement result from physical and/ or chemical interactions between the sensors and the volatile compounds pres- ent in the headspace above the measured sample. By using a QMB sensor sys- tem with an array of six sensors, good discrimination between three hop vari- eties can be observed (Fig. 15.10a). In this example only 12 measurements per sample were analysed. The distance between clusters is reduced if the data set is increased to 50 measurements per sample (Fig. 15.10b). The reason for this effect has to be attributed to a better and adequate ratio between sample size and array dimensionality. For a significant clustering of the patterns, with an array of six sensors a sample size of at least 18 is required [149, 184]. As a consequence, the discrimination based on only 12 measure- ments has poor statistical relevance. Most of the applications with sensor arrays found in the literature do not consider this fact; frequently discriminations with 12–32 sensors in an array and with a sample size of three to four are described. All of them are of limited feasibility with concurrent poor validation, especially in terms of reproducibility and predictive ability. In other words, if there are not enough calibration measurements one can separate data in a predetermined way, but will fail to verify the result using independent test samples. A great disadvantage of MOS, MOSFET, conducting polymer and QMB sen- sor arrays is that system-to-system matching is not possible in practice, as can

332 15 Advanced Instrumental Analysis and Electronic Noses be seen in Fig. 15.11. A discrimination of hop varieties with QMB and MOS arrays is partly obtained, but the results are not comparable. The hop variety Tradition, for example, is well separated from other varieties with the MOS sen- sor, whereas with the QMB sensor overlapping with other varieties is observed. The different responses and sensitivities of the sensors for chemical compounds is the reason for non-comparable discriminations when different sensor systems are used; therefore, a standardisation of these electronic noses is not possible. Fig. 15.10 Discrimination of hop varieties with six quartz crystal microbalance (QMB) sensors with 12 (a) and 50 (b) measurements per sample. N Nugget, S Select, M Magnum, P Perle, T Tradi- tion, B Northern Brewer Fig. 15.11 Discrimination of six hop varieties (see Fig. 15.10) by means of metal oxide sensor (MOS) and QMB sensor arrays

15.4 Electronic Noses 333 The interesting question now is whether the sensors classified the differ- ent hop samples on the basis of interaction with substances in the headspace which are highly correlated with the varieties. The GC analyses clearly show that myrcene is the main component in the headspace and is therefore responsible for the signals generated by the sensors. But the content of myrcene in hops is highly dependent on climatic, soil, growth and processing conditions, and can- not be regarded as a specific indicator for a variety assessment. Therefore, not the different hop varieties, but the different content of myrcene in the samples was discriminated. Frequently, authors tend to correlate an observed discrimination with the property they wanted to measure, without any additional chemical in- formation. Electronic noses as commonly employed do not allow for chemical differentiation. Owing to the unspecific nature of the sensors, the reasons for a successful discrimination of samples are usually unknown. In order to correlate a discrimination to the different varieties, a sensor sys- tem that selectively interacts with variety-specific compounds in the headspace is needed. The GC analysis of the essential oil reveals that there are some minor volatile compounds, which can be used for a differentiation of different hop va- rieties (e.g. Nugget and Tettnanger), as shown in Fig. 15.12. By choosing appropriate fragment ions a virtual sensor array based on a mass spectrometer in single ion monitoring mode can be implemented to discrimi- nate the hop varieties (Fig. 15.13) without appreciable interference of the main component myrcene. In contrast to the analyses performed with QMB and MOS sensors, the discrimination obtained with the MS sensor is based on chemical knowledge and not on assumptions. Fig. 15.12 Expanded region of a GC chromatogram of the hop varieties Nugget and Tettnanger

334 15 Advanced Instrumental Analysis and Electronic Noses Fig. 15.13 Discrimination of hop varieties with a mass spectrometry based sensor array 15.4.9 Conclusions The electronic nose technology applied to food must be regarded as being in its early stage. There is rapidly advancing research and development going on both for sensors and instrument hardware and software in order to enhance selectiv- ity, sensitivity and reproducibility of the gas sensors. Much effort is also being put into solving the drift problem of the sensors by increasing their stability and lifetime, and into developing improved mathematical algorithms for drift coun- teraction, automatic calibration and standardisation, as well as transferability between gas sensor array instruments. The performance of common multisensor arrays is ultimately determined by the properties of their constituent parts. Key parameters such as number, type and specificity of the sensors determine whether a specific instrument is suit- able for a given application. The selection of an appropriate set of chemical sen- sors is of utmost importance if electronic nose classifications are to be utilised to solve an analytical problem. As this requires time and effort, the applicability of solid-state sensor technology is often limited. The time saved compared with classic analytical methods is questionable, since analysis times of electronic nose systems are generally influenced more by the sampling method utilised than the sensor response time [185]. Common electronic noses are so called as they are often aimed at detection of odorous compounds; it is generally not clear that discriminations are based on odorous rather than non-odorous, and possibly incidental, components of the headspace. In the headspace of a food sample, odorants contributing to the flavour may be present in low concentrations, whereas non-odorous molecules can be present in much larger numbers and higher concentration. In such cases,

15.4 Electronic Noses 335 Table 15.3 Selected applications of electronic nose systems to different food products Product Type of application Sensor technology References Grains Off-odour caused MOS [107] by microbial infection MOSFET, MOS, IR [108, 109] CP [110] Meat Quality assessment, lipid MOSFET, MOS [111, 112] oxidation, fermentation, MOSFET, MOS, IR [110, 103] storage spoilage/shelf life CP [114, 115] CO, H2S, SO2, NH3 [99] Fish Quality assessment, MOSFET [116] lipid oxidation, spoil- MOSFET, MOS [117] age, freshness, storage MOS [118–120] CP [121, 122] CO, H2S, SO2, NH3 [98, 123] QMB [123, 124] IMS [97] Dairy Flavour quality, cheese BAW [125] products characterisation, heat MOS [126, 131–133, 139] treatment, flavour CP [127, 129, 134, 135], [137, differences, off flavours, MOS, CP, QMB 138] microbial contaminants MOS, CP [128, 130, 136, 140–144] MS Fruits Flavour quality, QMB [145] harvest dates, storage, MOS, MS [146] maturity, processing MOS [147–149] CP [150] Alcoholic Wines, spirits, origin, MOS [151] beverages variety, barrel age- MS [152–155] ing, cork taint MOS, CP, QMB [156] Beer Characterisation MS [157] and hops of aroma, ageing, raw MOS [158] materials, hop varieties CP [159, 160] QMB, MOS, MS [149] Spices Characterisation, dif- MOSFET, MOS [161] ferentiation, composition CP [162] of mixtures, microencap- MOS [163, 167, 168] sulation, γ-irradiation QMB [93, 164, 165] MS [94, 166] Olive oil Oxidation, rancidity, vin- MOSFET, MOS [169] egary defects, distinguish SAW [170] different qualities, shelf MOS [171, 172] life, geographical origin? CP [173] CP, MOS, MOSFET [174] MS [175] Coffee Discrimination, roasting CP [176] MOS [177, 178, 180] REMPI-TOFMS [179] Packaging Retained solvents, MOS, CP, QMB [181] printing inks, colouring QMB [182] agents, foil adhesives MOSFET, MOS, IR [183] QMB, MOS, MS [149] MOS metal oxide sensor, MOSFET metal oxide semiconductor field-effect transistor, IR infrared, CP conducting polymer, QMB quartz crystal microbalance, IMS ion mobility spectrometry, BAW bulk acoustic wave, MS mass spectrometry, SAW surface acoustic wave, REMPI-TOFMS reso- nance-enhanced multiphoton ionisation time-of-flight mass spectrometry

336 15 Advanced Instrumental Analysis and Electronic Noses MS-based systems have considerable advantages over the commonly used gas- sensor arrays, particularly in terms of selectivity, adaptability, sensitivity and standardisation. Array selection and deselection can be done rapidly by chang- ing the scanning method and/or simply by changing the fragment ions used for pattern recognition. Furthermore, taking into account that fragment ions contain chemical information about the sample, the information that can be obtained with an electronic nose improves substantially with the MS-based de- vices. An optimal instrument configuration also allows the same instrument to be used as a rapid screening tool (electronic nose) and also as a research tool for revealing further chemical information about doubtful samples. A great advan- tage is obtained with soft ionisation methods like PTR-MS or REMPI because molecular information is easier to deconvolute in the case of overlapping frag- ments or parent ions. Although common electronic noses are generally not suitable for odour as- sessment, they can be successfully used in applications where the main com- ponents in the headspace are directly correlated with the property to be deter- mined (e.g. quality of spice mixtures) or in cases where substances are formed and released into the headspace, for example owing to oxidation processes, fer- mentation, microbial contamination, thermal treatment, etc. 15.5 Time-Resolved Analysis of Volatile Organic Compounds Over the last decade, interest in release and delivery of VOCs has been steadily growing, with a particular focus on food, environmental and medical applica- tions [186–190]. Consequently, considerable effort was invested to develop ana- lytical methods capable of capturing such dynamic VOC release processes (Fig. 15.14) [179, 191]. This led to improvements in electronic sensor methods (often termed “electronic noses”) [192]. One other approach is direct-inlet MS. A prerequisite for mass analysis is ionisation, a process that critically influences the sensitivity and selectivity of the experiment. Electron impact ionisation (EI) causes considerable fragmentation. Because of overlapping fragment and parent ions, the molecular information is difficult to deconvolute, and little chemical information can be extracted. Therefore, application of direct-inlet MS for monitoring complex mixtures of VOCs requires using ionisation techniques which produce little or no fragmen- tation (soft ionisation). Chemical ionisation in combination with a quadrupole mass filter, either in atmospheric pressure chemical ionisation MS (APCI-MS) [188, 189] or in PTR-MS [193–195], have been successfully applied to VOC analyses. The advantages and limitations of direct-inlet MS with soft-ionisation approaches have been discussed [196]. One particularly well-performing technique is PTR-MS [193–195]. On-line trace-gas analysis by proton transfer [197] has become a powerful approach, mainly owing to the higher sensitivity and lower ionisation-induced fragmen-

15.5 Time-Resolved Analysis of Volatile Organic Compounds 337 Fig. 15.14 Analytical techniques for time-resolved headspace analysis. An electronic nose can be used as a low-cost process-monitoring device, where chemical information is not mandatory. Elec- tron impact ionisation mass spectrometry (EI-MS) adds sensitivity, speed and some chemical infor- mation. Yet, owing to the hard ionisation mode, most chemical information is lost. Proton-trans- fer-reaction MS (PTR-MS) is a sensitive one-dimensional method, which provides characteristic headspace profiles (detailed fingerprints) and chemical information. Finally, resonance-enhanced multiphoton ionisation (REMPI) TOFMS combines selective ionisation and mass separation and hence represents a two-dimensional method. (Adapted from [190]) tation relative to EI, the latter preventing efficient on-line trace-gas analysis of volatile mixtures via direct MS. In contrast, fragmentation by EI-MS is advanta- geous if used as a detector for GC to unequivocally identify pure compounds. An alternative to chemical ionisation is resonant (and non-resonant) laser ionisation methods [179], i.e. selective and soft laser photoionisation, such as REMPI. A particularly interesting setup is the combination of REMPI with TOFMS for monitoring coffee brew headspace. This chapter deals with tech- nical features and applications of time-resolved analytical methods with par- ticular focus on PTR-MS and resonant and laser ionisation methods (REMPI- TOFMS). 15.5.1 Proton-Transfer-Reaction Mass Spectrometry PTR-MS combines a soft, sensitive and efficient mode of chemical ionisation, adapted to the analysis of trace VOCs. Briefly, headspace gas is continuously introduced into the chemical ionisation cell, which contains besides buffer-gas a controlled ion density of H3O+. VOCs that have proton affinities larger than water (proton affinity of H2O is 166.5 kcal/mol) are ionised by proton transfer from H3O+, and the protonated VOCs are mass-analysed. The chemical ioni- sation source was specifically designed to achieve versatility, high sensitivity and little fragmentation, and to allow for absolute quantification of VOCs. To

338 15 Advanced Instrumental Analysis and Electronic Noses achieve these targeted specifications, the generation of the primary H3O+ ions and the chemical ionisation process—proton transfer from H3O+ to VOCs—are spatially and temporally separated and individually controlled. This allows (1) maximising signal intensity by increasing the generation of primary reactant ions, H3O+, in the ion source, (2) reducing fragmentation and clustering by op- timising the conditions for proton transfer in the drift tube and (3) quantifying VOCs from measured count rates. The four key features of PTR-MS can be summarised as follows. First, it is fast. Time dependent variations of headspace profiles can be monitored with a time resolution of better than 1 s. Second, the volatile compounds do not expe- rience any work-up or thermal stress, and very little fragmentation is induced by the ionisation step; hence, measured mass spectral profiles closely reflect genuine headspace distributions. Third, measured mass spectral intensities can be directly related to absolute headspace concentrations, without calibration or use of standards. Finally, it is not invasive and the process under investigation is not affected by the measurements. All these features make PTR-MS a particu- larly suitable method to investigate fast dynamic process. 15.5.1.1 Technical Features PTR-MS was introduced in 1993 by Lindinger and co-workers at the university of Innsbruck. A schematic drawing of the apparatus is given in Fig. 15.15. Here, only a brief description will be given. A more detailed discussion of the technical aspects of PTR-MS has been published in a series of review papers [193–195]. Primary (reactant) ions A+, generated in a hollow cathode ion source, travel through a buffer gas within the drift tube, to which the reactant gas (VOC) is added in small amounts, so that the density of the buffer gas is much larger than the density of the VOC. On their way through the reaction region, ions perform many non-reactive collisions with buffer gas atoms or molecules; however, once they collide with a reactant gas particle, they may undergo a reaction: (1) When H3O+ is used as the proton donor, most of the organic trace compo- nents R in air are ionised by proton-transfer processes: (2) These reactions are invariably fast, whenever they are exoergic, with rate co- efficients, k, close to the collisional limiting values, k0 ≈ 10-9 cm3/s- [197]. Water has a proton affinity of 7.22 eV (166.5 kcal/mol), and common organic mole-

15.5 Time-Resolved Analysis of Volatile Organic Compounds 339 cules have proton affinities in the range from 7 to 9 eV (161–208 kcal/mol), as shown in Table 15.4. Hence, most of the relevant proton-transfer reactions involving H3O+ are slightly exoergic, and H3O+ will perform proton-transfer reactions with nearly any kind of VOC in the headspace of food products. However, H3O+ does not react with the natural components of air such as O2, N2, CO2, CO or others (see Table 15.4). The exoergicity of the proton-transfer reaction with most VOCs, however, is low enough that breakup seldom occurs. On the basis of this ionisa- tion principle, a PTR-MS setup was developed applicable to trace-gas analysis, and aimed at speed, sensitivity, versatility and simple handling. The example shown in Fig. 15.16 was obtained by reconstituting a powdered beverage with hot water while measuring the headspace VOCs on-line by PTR- MS. It shows the relative ratio of the compounds released into the headspace and their dynamic behaviour. However, it is hardly possible to assign the mass traces to individual VOCs. For that, coupling of PTR-MS with GC-MS is re- quired, which will be discussed in the next section. Fig. 15.15 The PTR-MS apparatus. It consists of a series of three main chambers. In the first cham- ber, H2O is introduced and protonated in an electrical discharge to form H3O+. These ions are then driven by a small field through an orifice into the drift tube (chemical ionisation chamber). Coaxial to this orifice, neutral volatile organic compounds (VOCs) are introduced into the drift tube and collide at thermal energies with H3O+. VOCs with proton affinities exceeding 166.5 kcal/mol are ionised by proton transfer from H3O+ and are accelerated out of the drift tube into the quadrupole mass filter and onto the detector. (Adapted from [190])

340 15 Advanced Instrumental Analysis and Electronic Noses Table 15.4 Proton affinities of the constituents of clean air and of various volatile organic com- pounds. All volatile organic compounds with a higher proton affinity than H2O (166.5 kcal/mol) will be protonated with a very high efficiency when colliding with H3O+. This is the case for most of the volatile organic compounds in the headspace of coffee, with the exception of the natural constit- uents of clean air. In contrast, if NH+4 is used as a chemical ionisation agent, only compounds with a proton affinity exceeding 204.0/kcal · mol are ionised (below dotted line). (Adapted from [190]) Compounds Proton affinities Name Formula (kcal/mol) He 42.5 Ne 48.1 Ar 88.6 O2 100.9 N2 118.2 CO2 130.9 CH4 132.0 N2O 136.5 CO 141.9 Water H2O 166.5 Butane C4H10 163.3 Hydrogen sulphide H2S 170.2 Hydrogen cyanide HCN 171.4 Formic acid HCOOH 178.8 Propane C3H6 179.8 Benzene C6H6 181.9 Methanol CH3OH 181.9 Acetaldehyde CH3COH 186.6 Acetonitrile CH3CN 188.0 Ethanol C2H5OH 188.3 Furane C4H4O 192.2 2,3-Butanedione C4H6O2 194.8 Acetone CH3COCH3 196.7 2,3-Methylbutanal C5H10O ~195 Ammonia NH3 204.0 Pyrrole C4H5N 207.6

15.5 Time-Resolved Analysis of Volatile Organic Compounds 341 Table 15.4 (continued) Proton affinities of the constituents of clean air and of various volatile or- ganic compounds. Compounds Proton affinities Name Formula (kcal/mol) Oxazole C3H3NO 208.2 Pyrazine C4H4N2 209.0 Pyrazole C3H3N2 212.8 Dimethylamine C2H7N 217.0 Pyridine C5H5N 220.8 Trimethylamine C3H9N 225.1 Fig. 15.16 Ion traces (in m/z) of VOCs released upon reconstitution of an instant beverage and analysed on-line by PTR-MS 15.5.1.2 Coupling of Proton-Transfer-Reaction Mass Spectrometry with Gas Chromatography–Mass Spectrometry The success of PTR-MS triggered interest in further improving its performance. Indeed, PTR-MS is a one-dimensional technique, and ions from a complex headspace, e.g. coffee, can often only be tentatively assigned. Ions from different compounds (parent and fragment ions) can overlap in PTR-MS and prevent an unambiguous identification of VOCs in a complex mixture [198]. There-

342 15 Advanced Instrumental Analysis and Electronic Noses fore, the attempt has been made to address this problem and propose an exten- sion of PTR-MS which allows for an unambiguous identification of headspace compounds. This is achieved by coupling GC with simultaneous PTR-MS and EI-MS detection. In this chapter, we can only introduce the basic features of the new setup: a technical and analytical extension of PTR-MS which removes this shortcoming, while preserving its salient and unique features. Combining separation of VOCs by GC with simultaneous and parallel detection of the GC effluent by PTR-MS and EI-MS, an unambiguous interpretation of complex PTR-MS spectra becomes feasible. A more detailed description of characteristic performance parameters, such as resolution, linear range and detection limit, has been published in a recent paper [199]. As an example, the novel setup was applied to the characterisation of coffee headspace as a complex food system Basically, an aliquot of the headspace is trapped in defined time periods on several Tenax® adsorbents for characteri- sation by GC-MS. Figure 15.17 shows the simultaneously recorded total ion counts of the EI-MS (top frame) and PTR-MS (bottom frame) for VOCs trapped on the first Tenax® cartridge. The GC-separated pure compounds are identified Fig. 15.17 Simultaneous EI-MS (top trace) and PTR-MS (bottom trace) total ion count analysis of coffee headspace. Identification was based on MS spectra obtained at 70 eV and the retention index of the reference compounds. (Adapted from [199])

15.5 Time-Resolved Analysis of Volatile Organic Compounds 343 by comparison of their EI-MS fragmentation patterns with the Wiley database (Wiley 7th edition) as well as their retention indices obtained with reference compounds. The PTR-MS spectrum allows the PTR-MS fragmentation pattern of the GC-separated pure compounds to be identified. GC traces over the entire 40 min of the GC run are shown in Figs. 15.18 and 15.19, for the compounds desorbed from Tenax® cartridge no. 1 (trapping time window between 1 and 3 min). The data reveal that the PTR-MS ion signal at m/z 111 is a superposition of ions originating from two different compounds, i.e. 2-acetylfuran and 5-methylfurfural, contributing with 29 and 71%, respec- tively, to the total ion peak intensity at m/z 111. Similarly, the PTR-MS ion signal at m/z 87 is a superposition of 57% 2-methyl-1-propanal and 39% 2-butanone, with traces from 4-methyl-2-pentanone and 2-methyl tetrahydrofuran-3-one (2% each). While the single PTR-MS traces shown in Fig. 15.18 represent a su- perposition of several compounds, a series of PTR-MS ion traces are shown in Fig. 15.19 that are nearly pure (more than 89%), indicating that essentially only one single compound contributes to the ion signal (with only traces from other VOCs). Hence, in an on-line PTR-MS measurement of coffee headspace, the ion masses at m/z 68, 75, 80 and 95 can be assigned to pyrrole, acetol, pyridine and Fig. 15.18 Unambiguous identification of the molecules assigned to the trace ions. This identifica- tion is only valid for the first 120-s period of Tenax® trapping. (Adapted from [199])